This project continues surveillance for cancer incidence on a prospective cohort of 41,400 Hawaii residents who were interviewed between 1975 and 1980. The subjects constitute a random sample Of the population and include all main ethnic groups in Hawaii. Food frequency information, smoking and alcohol histories, height, weight, occupation, and demographic data were collected on each person. On a small subset of the cohort (4,890 persons), a more extensive, quantitative diet history was obtained. Because we have shown that out-migration rates in this cohort are extremely low, surveillance can be carried out entirely by computer linkage with the statewide, population-based Hawaii Tumor Registry. By the end of 1996 (fourth year of the renewal period), we expect the following numbers of cases of the cancers of interest: 544 lung, 360 colon, 531 breast, 410 prostate, 187 stomach, 116 pancreas, 82 oropharynx, 142 bladder, and 100 endometrium. Two specific aims will be pursued in the renewal period: 1) The relationship of antecedent diet to cancer occurrence at specific sites will be investigated. This analysis will include additional cancers that could not be studied previously because of small numbers of cases (oropharynx, stomach, pancreas, bladder, and endometrium). For the more common cancer sites (lung, colon, breast, and prostate), analyses will be extended to include ethnic-specific comparisons and comparisons by place of birth. These analyses will utilize recently-developed subgroup-specific portion size estimates. In addition, further assessments of the quality of the questionnaire data will be made, utilizing the more complete information on the special subset of the cohort. 2) Total mortality and competing risks of death will be examined in relation to dietary and anthropometric risk factors. To test the dietary hypotheses (aim 1), we will use proportional hazards regression to compare incidence among exposed and non-exposed cohort members, while adjusting for potential confounders. Examples of relationships to be examined for the less common sites include: meat intake and pancreatic cancer risk; salted fish and pickled vegetable consumption and stomach cancer risk; and fresh fruit and raw vegetable consumption and oropharyngeal, stomach, pancreatic, bladder, and endometrial cancer risk. To evaluate the effects of dietary and anthropometric factors on total mortality (aim 2), we will use the proportional hazards model to regress the time-to-death on the covariate of interest, after adjusting for confounding variables. Using this method, we will also examine the effects on mortality of various degrees of adherence to the recommended dietary guidelines from NCI and other agencies. A major strength of this cohort is the efficiency and low cost with which the surveillance and analyses can be carried out to yield useful information on associations between dietary components and cancer risk.